Towards a reliable animal model of migraine

Migraine used to be a favourite topic of the pharmaceutical industry. A number of triptans were developed and most of them successfully reached the market and had good or decent sales. After the triptan wave many companies have, however, had disappointments. Increasingly, the pharmaceutical industry is averse to migraine drug development. Even validated targets for drug development are not pursued by the industry. To change this pattern, an understanding is needed of the limited efficacy of the triptans and of the need for new treatments, as previously described by us (1). However, it is equally important to develop a reliable animal model suitable for the testing of new anti-migraine drugs. Many models exist, such as the dural extravasation model (2,3), the sagittal sinus stimulation model (4,5), the closed cranial window model (6), and various models recording single cell activity in the brainstem (7–12). While these have all contributed to a better understanding of the possible mechanisms of migraine, none of them are sufficiently predictive and versatile when it comes to the development of new drugs. This is particularly true for the development of new prophylactic drugs, the indication that is most in need of more specific and more effective drugs.

By far the best validated and most studied human migraine model uses intravenous infusion of nitroglycerin, also called glyceryl trinitrate (GTN). It is now generally accepted that infusion of GTN induces migraine attacks in migraineurs that are indistinguishable from spontaneous attacks (13). In normal volunteers without migraine, GTN induces a somewhat milder headache which has some but not all characteristics of migraine (14). Thus, normal persons rarely or never have the delayed migraine-like headache seen in migraine sufferers. The question is therefore whether an animal model is possible given the fact that animals probably do not suffer from migraine. A fairly large number of attempts have been made to transfer the GTN model to animals. Most of these studies have used rats and most of them have been done in anaesthetized animals (15–17) or have used intraperitoneal injection of GTN (18–24). Unfortunately, the latter route of administration seems to require enormous doses of GTN in order to elicit changes in the brain that may be compatible with migraine. Thus, approximately 1000 times the dose used in human experiments has been used in most such experiments. There are other problems with these studies. GTN does not dissolve in saline and usually must be dissolved in a mixture of alcohol and propylene glycol. Many studies using intraperitoneal GTN have, however, used saline for control and not vehicle (18,25–28). Because very large amounts of GTN are necessary, the amount of injected alcohol and propylene glycol are considerable. Furthermore, it is unknown what the effect of alcohol and propylene glycol into the peritoneum might be. In all likelihood it is painful, while saline used as control is not. Hence, the specificity of the responses obtained in most of these animal experiments is questionable. The huge doses of GTN induce a marked and prolonged blood pressure decrease in rats (29). Although it has been reported that i.p. injection of 10 mg/kg GTN has no effect on blood pressure in mice (18) we believe this may be questioned. Human use of GTN causes a markedly increased heart rate and occasionally vaso-vagal attacks while no increase in heart rate was reportedly seen in mice after 1000 times the human dose (18,30). It has been suggested that the enormous dose of GTN is necessary because rat liver is less efficient than human liver in transforming GTN to NO (18). However, this transformation does not take place only in the liver but also in tissues throughout the body, including vascular smooth muscle (31,32). Thus, GTN causes vasodilatation in isolated blood vessels. Other explanations must therefore be sought for the need of such enormous doses.

We have recently presented a modification of the GTN model in rats (33). The aim was to develop a model that resembles the human model as closely as possible. First and foremost, the animals had indwelling catheters and they recovered from surgery for 2 weeks before the experiments. Thus, surgical experimental pain and stress was avoided. Secondly, the animals were not anaesthetised, similar to the human situation. Vehicle was, of course, used for control not saline. Using this methodology it was possible to induce changes in migraine-relevant tissues using a dose that was only six times the human dose. This is generally considered to be the rat equivalent as rats always need a higher dose per kilogram than humans. Using this approximation to the human model, a clear response to sumatriptan treatment was seen. Clearly, this model is much more naturalistic than previous animal models but it is not the answer to all our wishes. GTN-induced headache in normal human volunteers did not respond to triptan treatment or aspirin (34). This indicates that NO in humans induces headache very deep in the cascade of events at a location not influenced by anti-migraine treatments. It may raise the question, therefore, why sumatriptan was actually effective in the rat model. In order to validate the rat model further, it would be necessary to assure the efficacy of 5-HT_1F receptor agonists and CGRP receptor antagonists. It might be that GTN is not at all the right provoking substance, and future studies should focus on the effect of other signalling molecules proven to induce migraine in human patients. Among the most interesting candidates are CGRP and PACAP.

Models that involve killing the animal are expensive and slow because an animal can only be used to demonstrate the effect of one dose at one particular time point. What would really move the field forward would be a behavioural model in which animals can be used again and again. With a reliable behavioural animal model, crossover experiments could be done in the same animal using placebo and one or more active treatments. There are models that perhaps will fulfil these requirements. In rats the ultrasonic vocalisation response to air puff has been suggested (35), but the method involves application of lipopolysaccharide on the dura mater to induce dural inflammation and it has never gained widespread use.

In transgenic mice carrying an FHM mutation, increased sensitivity seems associated with a particular facial expression and behaviour perhaps suggesting head pain (36). Unfortunately, FHM patients do not have increased sensitivity to GTN or CGRP and the mouse model harbouring this mutation may not be relevant to the common types of migraine: migraine without aura and migraine with typical aura. Further studies of the ability of the transgenic models to predict anti-migraine efficacy are needed.

In conclusion, many attempts have been made to develop animal models predictive of efficacy of anti-migraine drugs. We are not there yet but development of such a model seems to be one of the most important avenues of headache research today. With a validated animal model at hand it is expected that the pharmaceutical industry will return to the migraine field and do much more to develop specific and more efficacious drugs for migraine.